CN109325995A - Low resolution multi-angle of view hand method for reconstructing based on manpower parameter model - Google Patents
Low resolution multi-angle of view hand method for reconstructing based on manpower parameter model Download PDFInfo
- Publication number
- CN109325995A CN109325995A CN201811071125.8A CN201811071125A CN109325995A CN 109325995 A CN109325995 A CN 109325995A CN 201811071125 A CN201811071125 A CN 201811071125A CN 109325995 A CN109325995 A CN 109325995A
- Authority
- CN
- China
- Prior art keywords
- manpower
- model
- hand
- parameter
- angle
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
Low resolution multi-angle of view hand method for reconstructing based on manpower parameter model, comprising: the multi-view image of acquisition human body entirety；Exposure mask, two-dimentional artis are extracted, three-dimensional skeleton point is converted into；Movement deformation is carried out to manpower parameter model, obtains manpower movement distorted pattern；The form parameter for optimizing manpower parameter model, obtains human hand shape's distorted pattern；Skeleton point and model back projection obtain final hand model to each multi-view image, the existing two-dimentional skeleton point of optimization update and exposure mask；Splice with the manikin.Whole process of the invention is automatically performed, and expresses gesture motion and shape feature is accurate, and reconstruction precision is high, and hand model and manikin are carried out splicing recombination, obtain more acurrate complete modelling effect.
Description
Technical field
The present invention relates to the manpower three-dimensional reconstruction fields in computer graphical middle school, specifically a kind of to be based on manpower parameter
The low resolution multi-angle of view hand method for reconstructing of model.
Background technique
The human body three-dimensional reconstruction technique of high quality is generally required based on complicated multi-angle of view camera system or depth sensing
Device system.But either color camera or depth transducer, when the gesture motion of people is slightly complicated, under this reconstructing system just
It will appear the situation more than fingertip tracking or hand noise, this is the significant challenge of complete acquisition manikin, and this method can
To increase substantially the reconstruction precision of hand model, recover high quality while not influencing manikin acquisition setting
Hand model is merged with human body.
For color camera, multi-angle of view human body reconstruction needs the Feature Points Matching based on picture, when human body completely goes out
When present entire picture, the region of hand often only accounts for the ratio of " very little "；And too low resolution ratio and originally similar hand
Portion's feature can all influence its Feature Points Matching result.For depth transducer, when human body is by its complete capture, according to human body
Height difference and sensor field of view angular difference, people need 3~4 meters of range sensor or more, common depth transducer, such as Kinect
Also there was only 0.5~4.5 meter of valid analysing range；Meanwhile the geometry of hand is complicated and fine, slightly deviation can result in
Much noise is lost and introduced to finger.Manpower joint is flexible, and movement is complicated, the reconstruction process of view-based access control model, can by hand itself
The influence mutually blocked with both hands is blocked, and then increases the difficulty that complicated gesture High-precision human is rebuild.
CN201810140587.4 patent application discloses a kind of three-dimensional manpower 3D based on binocular color imaging system
Skeleton pattern real-time reconstruction method, binocular color imaging system obtain two the same hand different perspectivess for shooting manpower
Color image；Method includes: detection hand position；Detect hand plane key point；According to the plan-position of key point in binocular image
Restore three-dimensional coordinate；Three-dimensional manpower skeleton is fitted using the three-dimensional coordinate of the key point.Binocular cromogram is used only in this method
Picture can opponent's progress 3D reconstruction.
CN201110235370X patent discloses a kind of three-dimension gesture movement method for reconstructing and system, comprising: to acquisition
Gesture image sequence first frame image carry out region segmentation；Generate the affine transformation matrix of segmentation area；It generates three-dimensional
Three-dimensional corresponding with the first frame image is obtained according to the projection coefficient and the affine transformation matrix being somebody's turn to do to two-dimensional projection's coefficient
Gesture model；To obtained three-dimension gesture model corresponding with the first frame image, its skeleton node and the skeleton node are determined
Freedom degree；For subsequent each frame image, skeleton node based on three-dimension gesture model corresponding with previous frame image and should be from
By spending, the operation of simulated annealing particle filter is carried out in conjunction with current frame image, obtains the three-dimension gesture model of current frame image, thus
Realize the reconstruction to three-dimension gesture.
Summary of the invention
The present invention is existing to solve the problems, such as, it is desirable to provide a kind of low resolution multi-angle of view hand based on manpower parameter model
Portion's method for reconstructing.
In order to achieve the above object, the technical solution adopted by the present invention includes the following steps:
Step 1 constructs manpower parameter model；The multi-view image of human body entirety is acquired, and establishes manikin；
Step 2 extracts the exposure mask of manpower in multi-view image, two-dimentional artis, and converts thereof into three-dimensional skeleton point；
Step 3 carries out movement deformation to manpower parameter model according to three-dimensional manpower skeleton point, obtains manpower movement deformation
Model；
Step 4 is based on the exposure mask, deformed hand model, optimizes the form parameter of manpower parameter model, obtains
Human hand shape's distorted pattern；
Step 5, by the skeleton point and model of human hand shape's distorted pattern, back projection to each multi-view image, optimization is more
New existing two-dimentional skeleton point and exposure mask, obtain final hand model；
Step 6 splices final hand model and the manikin.
Further, in step 2, three-dimensional is calculated with triangulation, boundling adjustment algorithm and skeleton physical constraint
Skeleton point X, formula are as follows:
argmin(Eproj+αEpose)
Wherein, p indicates key point number；V indicates camera perspective number；Indicate the camera perspective of the visible manpower of f frame
Set；For the confidence level of depth network output；Pv() is projection function.
Further, in step 3, the transformation square of entire palm is calculated according to the position distribution of the three-dimensional skeleton point of palm
Battle array, including displacement, rotation, scaling；Shift transformation and the rotation transformation for successively calculating each finger skeleton point again, obtain people
Make distorted pattern manually.
Further, in step 4, distorted pattern is acted according to the exposure mask score chart of the manpower of multi-view image and manpower,
Utilize following multi-angle of view projection optimization equation:
Optimize the form parameter of manpower parameter model, makes one to make exposure mask and input of the distorted pattern under each visual angle manually
It is close, and then obtain human hand shape's distorted pattern；
Wherein, V: camera perspective set in one's hands is detected；NV: gather number in the visual angle；S: shape optimum parameter it is initial
Value；S': the target value of shape optimum parameter；Ns: deformation parameter number；P': based on the vertex position after s' deformation；Gi: model shape
Visible p' vertex set under i-th visual angle after change；NGi: corresponding GiGather number of vertices；Ii m: the manpower exposure mask point under the i-th visual angle
Number figure；Take the score of the exposure mask score chart corresponding position；Mi: the projection matrix at the i-th visual angle；lp': corresponding vertex p' institute
The finger classification belonged to；wj: the confidence level of the point of three-dimensional framework corresponding to j-th of deformation parameter.
Further, in step 5, the two-dimentional artis and exposure mask of each multi-angled view of Reverse optimization deform human hand shape
The three-dimensional skeleton point X of model projects to each visual angle and then obtains two-dimentional skeleton point x, generates bone site figure and original picture using x
Be input to together artis detection depth network in, conjugation condition random field optimized after two-dimensional framework point position；It will
Exposure mask is inputted together with original picture in the depth network of semantic segmentation, is assisted optimizing also with condition random field, be optimized
The exposure mask at each visual angle afterwards.
It further, will be Step 2: three and four iterations, until the knot of hand model parameter optimization after step 5
Fruit variation then stops when little, obtains final hand model.
Further, in step 6, the deep neural network detected first according to human joint points, to extract each visual angle figure
As upper two-dimentional artis position；Then restore three-dimensional skeleton point position；Secondly using three-dimensional elbow skeleton point position, driving is most
The position of elbow skeleton point in whole hand model；Finally by already present hand in final hand model and manikin and forearm into
The accurate matching alignment of row, removes already present hand region in manikin, by final hand model and human body model splicing
Together.
Compared to the prior art, the present invention is based on people's gymnadenia conopsea exponential model of setting, Auto-matching parameter model and bats
Action parameter, the form parameter of object hand are taken the photograph, to replace the hand region of original manikin；Make full use of the figure at each visual angle
Piece ensure that sufficient human body information input；The driving that hand model is not only carried out using the critical point detection of image, is also tied
Semantic segmentation algorithm is closed, the region of each finger on the image is partitioned into, the optimization of human hand shape's parameter is carried out, makes human hand shape more
Add precisely；
The present invention combines shape with movement, using manpower key point, manpower semantic segmentation result as input, successively follows
Ring optimised shape data, action data, and in optimization process, it is intermediate to two-dimentional artis, three-dimensional skeleton point, semantic segmentation etc.
As a result it is optimized；The target for specifying each step optimization process all has the data input of different accuracy good
Robustness；
Whole process of the invention is automatically performed, and expresses gesture motion and shape feature is accurate, reconstruction precision is high, by hand
Model and manikin carry out splicing recombination, obtain more acurrate complete modelling effect.
Detailed description of the invention
Fig. 1 is the schematic diagram of annular camera array；
Fig. 2 is the schematic diagram of people's gymnadenia conopsea exponential model；
Fig. 3 is the flow diagram of one embodiment of the present of invention.
Specific embodiment
The present invention is further described now in conjunction with attached drawing.
Referring to Fig. 1 to Fig. 3, Fig. 1 to Fig. 3 shows one embodiment of the present of invention.It is annular camera array in Fig. 1,
Including 48 cameras；The present embodiment is proceeded as follows using the equipment:
Step 1, data acquisition.One people is shot using camera array, the surrounding of people is green curtain background；By shooting
To the photo of this person under a multiple of viewing angles.Followed by the Feature Points Matching algorithm and boundling adjustment algorithm weight of multi-angle of view picture
It builds to obtain the poor manikin of hand precision.Feature Points Matching algorithm and boundling adjustment algorithm are the prior art, and non-
Invent institute's emphasis referents.
Referring to fig. 2, manpower parameter model PH (Parametric Hand) is set；The manpower parameter model, has human body
The general modelling feature of palm, finger；But it is a lack of the details of specific action and part.The people's gymnadenia conopsea exponential model can be according to need
It is acted, the change of local moulding.
Step 2, exposure mask and skeleton point obtain.The hand mask and two-dimentional bone on image are obtained using deep neural network
Bone point x position calculates more accurate human body three-dimensional bone with triangulation, boundling adjustment algorithm and skeleton physical constraint
Bone point X, formula are as follows:
argmin(Eproj+αEpose)
Wherein, p indicates key point number；V indicates camera perspective number；Indicate the camera perspective of the visible manpower of f frame
Set；For the confidence level of depth network output；Pv() is projection function.
Step 3, the action drives of manpower parameter model PH.Referring to fig. 2, it according to obtained manpower three-dimensional skeleton point X, drives
It makes one gymnadenia conopsea exponential model, calculates transformation matrix (including the position of entire palm according to the position distribution of palm three-dimensional skeleton point first
Move, rotation, scaling), then shift transformation and the rotation transformation of each finger skeleton point are successively calculated, it obtains manpower movement and becomes
Shape model AHM (Animated Hand Model).
Step 4, the form parameter optimization of manpower parameter model PH；According to the exposure mask score chart of the manpower of multi-view image
Distorted pattern is acted with manpower, utilizes following multi-angle of view projection optimization equation:
Optimize the form parameter of manpower parameter model, makes one to make exposure mask and input of the distorted pattern under each visual angle manually
It is close, and then obtain human hand shape's distorted pattern DHM；
Wherein, V indicates to detect camera perspective set in one's hands；NVIndicate that number is gathered at the visual angle；S indicates shape optimum ginseng
Several initial values；The target value of s' expression shape optimum parameter；NsIndicate deformation parameter number；After p' is indicated based on s' deformation
Vertex position；GiVisible p' vertex set under i-th visual angle after expression model deformation；NGi indicates corresponding GiGather number of vertices；
Ii mIndicate the manpower exposure mask score chart under the i-th visual angle；Expression takes the score of the exposure mask score chart corresponding position；MiIt indicates
The projection matrix at the i-th visual angle；lp'Indicate finger classification belonging to corresponding vertex p'；wjIt indicates corresponding to j-th of deformation parameter
Three-dimensional framework point confidence level.
Step 5, the two-dimentional skeleton point and exposure mask at each visual angle of Reverse optimization.By the three-dimensional bone of human hand shape's distorted pattern DHM
Bone point X projects to each visual angle and obtains two-dimentional skeleton point x, and it is defeated together with original picture to generate bone site figure using two-dimentional skeleton point x
Enter in the depth network detected to artis, conjugation condition random field, the two-dimensional framework point position after being optimized；By exposure mask with
Original picture is inputted together in the depth network of semantic segmentation, assists optimizing also with condition random field, each after being optimized
The exposure mask of angle.
Then, for iteration Step 2: three and four, number is two to more than three times；To the knot of hand model parameter optimization
Fruit variation then stops iteration when little, obtains the final hand model FHM optimized.
Final hand model FHM and original manikin are carried out splicing by step 6, specifically:
According to the deep neural network that human joint points detect, two-dimension human body artis position on each multi-perspective picture is extracted.
Using triangulation algorithm identical with people's calculating hand skeleton point, restore the position of three-dimensional skeleton point x.
Using three-dimensional elbow skeleton point position, the position of elbow skeleton point in final hand model FHM is driven, based on a cloud
Final hand model FHM is carried out accurate matching with forearm with hand already present in manikin and is aligned by iteration with regard to proximal point algorithm,
Already present hand region in manikin is removed, using Poisson surface algorithm for reconstructing by final hand model FHM and the human body
Model splicing together, obtains final fused manikin.It is a feature of the present invention that being based on people's gymnadenia conopsea digital-to-analogue
Type can be with the action parameter of Auto-matching parameter model and reference object hand, form parameter, to replace original manikin
Hand region.
Whole process is automatically performed, and expresses gesture motion and shape feature is accurate, reconstruction precision is high, by hand model and people
Body Model carries out splicing recombination, obtains more acurrate complete modelling effect.
Embodiments of the present invention are described above in conjunction with accompanying drawings and embodiments, the not composition that embodiment provides is to this hair
Bright limitation, those skilled in the art in the art can make within the scope of the appended claims according to needing to adjust
Various deformations or amendments are in protection scope.
Claims (7)
1. a kind of low resolution multi-angle of view hand method for reconstructing based on manpower parameter model, it is characterised in that including walking as follows
It is rapid:
Step 1 constructs manpower parameter model；The multi-view image of human body entirety is acquired, and establishes manikin；
Step 2 extracts the exposure mask of manpower in multi-view image, two-dimentional artis, and converts thereof into three-dimensional skeleton point；
Step 3 carries out movement deformation to manpower parameter model according to three-dimensional manpower skeleton point, obtains manpower movement distorted pattern；
Step 4 is based on the exposure mask, deformed hand model, optimizes the form parameter of manpower parameter model, obtains manpower
Shape distortion model；
Step 5, by the skeleton point and model of human hand shape's distorted pattern, back projection to each multi-view image, optimization updates
The two-dimentional skeleton point and exposure mask having, obtain final hand model；
Step 6 splices final hand model and the manikin.
2. a kind of low resolution multi-angle of view hand method for reconstructing based on manpower parameter model according to claim 1,
It is characterized in that: in step 2, calculating three-dimensional skeleton point X with triangulation, boundling adjustment algorithm and skeleton physical constraint,
Formula are as follows:
argmin(Eproj+αEpose)
Wherein, p indicates key point number；V indicates camera perspective number；Indicate the camera perspective set of the visible manpower of f frame；For the confidence level of depth network output；Pv() is projection function.
3. a kind of low resolution multi-angle of view hand method for reconstructing based on manpower parameter model according to claim 1,
It is characterized in that: in step 3, the transformation matrix of entire palm, including position is calculated according to the position distribution of the three-dimensional skeleton point of palm
It moves, rotation, scaling；Shift transformation and the rotation transformation for successively calculating each finger skeleton point again obtain manpower movement and become
Shape model.
4. a kind of low resolution multi-angle of view hand method for reconstructing based on manpower parameter model according to claim 1,
It is characterized in that: in step 4, distorted pattern being acted according to the exposure mask score chart of the manpower of multi-view image and manpower, using as follows
Multi-angle of view projection optimization equation:
Optimize the form parameter of manpower parameter model, makes one to make exposure mask of the distorted pattern under each visual angle manually to connect with input
Closely, and then human hand shape's distorted pattern is obtained；
Wherein, V: camera perspective set in one's hands is detected；NV: gather number in the visual angle；S: the initial value of shape optimum parameter；S':
The target value of shape optimum parameter；Ns: deformation parameter number；P': based on the vertex position after s' deformation；Gi: after model deformation
Visible p' vertex set under the visual angle i；NGi: corresponding GiGather number of vertices；Ii m: the manpower exposure mask score chart under the i-th visual angle；Take the score of the exposure mask score chart corresponding position；Mi: the projection matrix at the i-th visual angle；lp': belonging to corresponding vertex p'
Finger classification；wj: the confidence level of the point of three-dimensional framework corresponding to j-th of deformation parameter.
5. a kind of low resolution multi-angle of view hand method for reconstructing based on manpower parameter model according to claim 1,
It is characterized in that: in step 5, the two-dimentional artis and exposure mask of each multi-angled view of Reverse optimization, by the three of human hand shape's distorted pattern
Dimension skeleton point X projects to each visual angle and then obtains two-dimentional skeleton point x, generates bone site figure using x and inputs together with original picture
To artis detect depth network in, conjugation condition random field optimized after two-dimensional framework point position；By exposure mask and original
Picture is inputted together in the depth network of semantic segmentation, assists optimizing also with condition random field, each view after being optimized
The exposure mask at angle.
6. a kind of according to claim 1, low resolution multi-angle of view hand weight based on manpower parameter model described in 2,3,4 or 5
Construction method, it is characterised in that:, will be Step 2: three and four iterations, until the knot of hand model parameter optimization after step 5
Fruit variation then stops when little, obtains final hand model.
7. a kind of low resolution multi-angle of view hand method for reconstructing based on manpower parameter model according to claim 6,
It is characterized in that: in step 6, the deep neural network detected first according to human joint points, to extract two on each multi-view image
Tie up artis position；Then restore three-dimensional skeleton point position；Secondly using three-dimensional elbow skeleton point position, final hand mould is driven
The position of elbow skeleton point in type；Already present hand and forearm in final hand model and manikin are finally carried out accurate
With alignment, already present hand region in manikin is removed, by final hand model together with human body model splicing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811071125.8A CN109325995B (en) | 2018-09-13 | 2018-09-13 | Low-resolution multi-view hand reconstruction method based on hand parameter model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811071125.8A CN109325995B (en) | 2018-09-13 | 2018-09-13 | Low-resolution multi-view hand reconstruction method based on hand parameter model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109325995A true CN109325995A (en) | 2019-02-12 |
| CN109325995B CN109325995B (en) | 2022-11-25 |
Family
ID=65265305
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811071125.8A Active CN109325995B (en) | 2018-09-13 | 2018-09-13 | Low-resolution multi-view hand reconstruction method based on hand parameter model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109325995B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110348375A (en) * | 2019-07-09 | 2019-10-18 | 华南理工大学 | A kind of finger vena region of interest area detecting method neural network based |
| CN110796699A (en) * | 2019-06-18 | 2020-02-14 | 叠境数字科技（上海）有限公司 | Optimal visual angle selection method and three-dimensional human skeleton detection method of multi-view camera system |
| CN111832468A (en) * | 2020-07-09 | 2020-10-27 | 平安科技（深圳）有限公司 | Gesture recognition method and device based on biological recognition, computer equipment and medium |
| CN112465890A (en) * | 2020-11-24 | 2021-03-09 | 深圳市商汤科技有限公司 | Depth detection method and device, electronic equipment and computer readable storage medium |
| CN112907631A (en) * | 2021-02-20 | 2021-06-04 | 北京未澜科技有限公司 | Multi-RGB camera real-time human body motion capture system introducing feedback mechanism |
| CN113238650A (en) * | 2021-04-15 | 2021-08-10 | 青岛小鸟看看科技有限公司 | Gesture recognition and control method and device and virtual reality equipment |
| WO2021184933A1 (en) * | 2020-03-20 | 2021-09-23 | 华为技术有限公司 | Three-dimensional human body model reconstruction method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102262783A (en) * | 2011-08-16 | 2011-11-30 | 清华大学 | Method and system for restructuring motion of three-dimensional gesture |
| CN105469113A (en) * | 2015-11-19 | 2016-04-06 | 广州新节奏智能科技有限公司 | Human body bone point tracking method and system in two-dimensional video stream |
| US20180053309A1 (en) * | 2016-08-22 | 2018-02-22 | Seiko Epson Corporation | Spatial Alignment of M-Tracer and 3-D Human Model For Golf Swing Analysis Using Skeleton |
-
2018
- 2018-09-13 CN CN201811071125.8A patent/CN109325995B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102262783A (en) * | 2011-08-16 | 2011-11-30 | 清华大学 | Method and system for restructuring motion of three-dimensional gesture |
| CN105469113A (en) * | 2015-11-19 | 2016-04-06 | 广州新节奏智能科技有限公司 | Human body bone point tracking method and system in two-dimensional video stream |
| US20180053309A1 (en) * | 2016-08-22 | 2018-02-22 | Seiko Epson Corporation | Spatial Alignment of M-Tracer and 3-D Human Model For Golf Swing Analysis Using Skeleton |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110796699A (en) * | 2019-06-18 | 2020-02-14 | 叠境数字科技（上海）有限公司 | Optimal visual angle selection method and three-dimensional human skeleton detection method of multi-view camera system |
| CN110796699B (en) * | 2019-06-18 | 2024-03-01 | 叠境数字科技（上海）有限公司 | Optimal view angle selection method and three-dimensional human skeleton detection method for multi-view camera system |
| CN110348375A (en) * | 2019-07-09 | 2019-10-18 | 华南理工大学 | A kind of finger vena region of interest area detecting method neural network based |
| WO2021184933A1 (en) * | 2020-03-20 | 2021-09-23 | 华为技术有限公司 | Three-dimensional human body model reconstruction method |
| CN113496507A (en) * | 2020-03-20 | 2021-10-12 | 华为技术有限公司 | Human body three-dimensional model reconstruction method |
| CN111832468A (en) * | 2020-07-09 | 2020-10-27 | 平安科技（深圳）有限公司 | Gesture recognition method and device based on biological recognition, computer equipment and medium |
| WO2021120834A1 (en) * | 2020-07-09 | 2021-06-24 | 平安科技（深圳）有限公司 | Biometrics-based gesture recognition method and apparatus, computer device, and medium |
| CN112465890A (en) * | 2020-11-24 | 2021-03-09 | 深圳市商汤科技有限公司 | Depth detection method and device, electronic equipment and computer readable storage medium |
| CN112907631A (en) * | 2021-02-20 | 2021-06-04 | 北京未澜科技有限公司 | Multi-RGB camera real-time human body motion capture system introducing feedback mechanism |
| CN113238650A (en) * | 2021-04-15 | 2021-08-10 | 青岛小鸟看看科技有限公司 | Gesture recognition and control method and device and virtual reality equipment |
| CN113238650B (en) * | 2021-04-15 | 2023-04-07 | 青岛小鸟看看科技有限公司 | Gesture recognition and control method and device and virtual reality equipment |
| US11947729B2 (en) | 2021-04-15 | 2024-04-02 | Qingdao Pico Technology Co., Ltd. | Gesture recognition method and device, gesture control method and device and virtual reality apparatus |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109325995B (en) | 2022-11-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109325995A (en) | Low resolution multi-angle of view hand method for reconstructing based on manpower parameter model | |
| Plänkers et al. | Tracking and modeling people in video sequences | |
| CN108154550B (en) | RGBD camera-based real-time three-dimensional face reconstruction method | |
| CN108765328B (en) | High-precision multi-feature plane template and distortion optimization and calibration method thereof | |
| US11210804B2 (en) | Methods, devices and computer program products for global bundle adjustment of 3D images | |
| CN104992441B (en) | A kind of real human body three-dimensional modeling method towards individualized virtual fitting | |
| US6664956B1 (en) | Method for generating a personalized 3-D face model | |
| CN102663820B (en) | Three-dimensional head model reconstruction method | |
| CN107067429A (en) | Video editing system and method that face three-dimensional reconstruction and face based on deep learning are replaced | |
| CN104167016B (en) | A kind of three-dimensional motion method for reconstructing based on RGB color and depth image | |
| CN108053437B (en) | Three-dimensional model obtaining method and device based on posture | |
| CN110599540A (en) | Real-time three-dimensional human body shape and posture reconstruction method and device under multi-viewpoint camera | |
| CN111932678B (en) | Multi-view real-time human motion, gesture, expression and texture reconstruction system | |
| KR20180069786A (en) | Method and system for generating an image file of a 3D garment model for a 3D body model | |
| CN105279769B (en) | A kind of level particle filter tracking method for combining multiple features | |
| CN109242954A (en) | Multi-view angle three-dimensional human body reconstruction method based on template deformation | |
| CN104794722A (en) | Dressed human body three-dimensional bare body model calculation method through single Kinect | |
| CN103824326B (en) | Dynamic human body three-dimensional modeling method | |
| Fua et al. | Human body modeling and motion analysis from video sequences | |
| CN104463969B (en) | A kind of method for building up of the model of geographical photo to aviation tilt | |
| CN112907631B (en) | Multi-RGB camera real-time human body motion capture system introducing feedback mechanism | |
| CN105258680B (en) | A kind of object pose measuring method and device | |
| CN104376599A (en) | Handy three-dimensional head model generation system | |
| CN113393577B (en) | Oblique photography terrain reconstruction method | |
| Plankers et al. | Automated body modeling from video sequences |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |